請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/65016
標題: | 基於自適應波束成形超聲波平面波成像技術的開發與其電路設計 Development of Adaptive Beamforming Based Ultrasound Plane Wave Imaging and Its Circuit Design |
作者: | Ming Khuan Son 孫明均 |
指導教授: | 闕志達(Tzi-Dar Chiueh) |
關鍵字: | 自適應波束成形相干平面波複合(CPWC),最小方差波束成形,Wiener後置濾波器, adaptive beamforming,coherent plane-wave compounding (CPWC),minimum variance beamforming,Wiener post-filter, |
出版年 : | 2020 |
學位: | 碩士 |
摘要: | 近年來超音波成像已被廣泛地使用於現代醫學應用上。超音波成像是一種有效的診斷工具,在診所以及醫院中都具有良好的安全使用記錄。超音波成像與基於非電離輻射,與使用電離輻射的X射線或其他類型的成像系統相反,超聲波成像不會給患者帶來風險。在經典的超聲波掃描過程中,超聲波脈衝通過人體的關注區域傳輸。 然後對反向散射的回波信號進行波束成形,以產生影像。 波束成形在超音波成像中起著關鍵作用,不同的演算法會影響最終圖像的分辨率(Resolution)和對比度(Contrast)。
我們研究了醫學超聲波成像中現有的波束成形方法,並簡要回顧了最常見的波束成形技術,從標準DAS波束成形方法開始,到最知名的自適應波束成形技術,例如Minimum Variance (MV)。 在論文的前半段,我們會簡要地介紹現有的Data Compounding on Receive Minimum Variance(DCR-MV)演算法。我們也會對此演算法做分析並且與其它現有的演算法做比較。我們會使用IEEE IUS 2016 Plane-wave Imaging Challenge in Medical UltraSound : PICMUS [1]所提供的超音波影像以及評分方式來評估這些演算法的優劣。接下來我們會取出目前最好的演算法並提出改善此演算法的方法,讓超音波產生之圖像的分辨率(Resolution)和對比度 (Contrast)都會有所提升。 除演算法模擬外,本論文有針對開發的演算法進行電路的設計,並在開發beamforming engine。此電路可有效地實現上述幾種超聲波束形成方法。這個beamforming engine能將輸入的ADC data經由運算後回傳影像。目前在FPGA上開發的版本可支援四種不同的演算法,其中包含最經典的DAS演算法、DCR-MV演算法以及我們新開發的兩個演算法。除此之外,我們的硬體也支援不同大小矩陣的反矩陣運算。我們在硬體上所使用的反矩陣運算電路架構為systolic array。與一般設計不同的是我們可使用6x6的反矩陣電路透過data scheduling去解出11x11的反矩陣以減少硬體資源的使用。 In recent years, ultrasound imaging has been widely used in modern medical applications. Ultrasound imaging is an effective diagnostic tool and has a good safety record in clinics and hospitals. In contrast to X-rays or other types of imaging systems that use ionizing radiation, ultrasound imaging does not pose a risk to the patient. During a classic ultrasound scan, short acoustic pulses are transmitted through the area of interest of the human body. The backscattered echo signals are then beamformed to generate an image. Beamforming plays a key role in ultrasound imaging, and different algorithms will affect the resolution and contrast of the beamformed image. We have studied the existing beamforming methods in medical ultrasound imaging. We briefly reviewed the most common beamforming techniques, starting with standard delay and sum beamforming methods, to the most well-known adaptive beamforming techniques, such as Minimum Variance (MV) and Data Compounding on Receive Minimum Variance (DCR-MV). In the first half of this thesis, we briefly introduce the existing Data Compounding on Receive Minimum Variance (DCR-MV) algorithm. We also analyze this algorithm and compare it with other existing algorithms. We use the ultrasonic images and scoring methods provided by IEEE IUS 2016 Plane-wave Imaging Challenge in Medical Ultrasound: PICMUS[1] to evaluate the pros and cons of these algorithms. Next, we choose the best current algorithm and propose methods to improve this algorithm, so that the resolution and contrast of the beamformed image will be improved. In addition to algorithm simulation, we also design the circuits for the developed algorithms and develop a beamforming engine. This circuit can effectively implement the above-mentioned several ultrasonic beamforming methods. The beamforming engine will then process the input ADC data and generate output data for the beamformed image. The current version developed on FPGA can support four different algorithms, including the most classic DAS algorithm, DCR-MV algorithm and two newly developed algorithms. In addition, our hardware also supports inverse matrix operations on matrices of different sizes. The inverse matrix arithmetic circuit architecture that we used on hardware is a systolic array. Unlike the general design, we can use the 6x6 inverse matrix circuit to solve the 11x11 inverse matrix through data scheduling to reduce the use of hardware resources. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/65016 |
DOI: | 10.6342/NTU202000546 |
全文授權: | 有償授權 |
顯示於系所單位: | 電子工程學研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-109-1.pdf 目前未授權公開取用 | 8.25 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。